The untitled "Manatee Quiz Game" for Mote Marine Laboratories is a 10-question quiz game set in 4 stages of rising difficulty. The questions and answers are randomized, and there are 3 lifelines: 50/50, Ask a Manatee, and Change the Question. The game is played in front of a 65" TV at a full 4K resolution at 60 frames per second in 10-bit HDR. A bit overkill for a quiz game? Good! That sure makes things interesting! Let's dive into this journey together...
I previously had one of Mote's scientists (Dr. Hall) as an instructor in Ringling College. My first official project with Mote was 2115. What started off as a class project eventually made its way to an exhibit on permanent display in the aquarium. I learned so much from that project. The quiz project started off from an e-mail I received from a man named Evan B. (Assistant VP of Mote) that appeared in my junk folder. He loved the 2115 project so much, he wanted me to make another game. I check my junk folder every day now.
The original 2115 video, made in UDK at Ringling College of Art and Design (2012)
Mote wanted a game with manatees as part of their exhibit. A couple ideas we bounced included a manatee-counting game (similar to aerial surveys taken by field researchers), and a quiz game. I was able to make a working prototype of a quiz game that shuffled the questions and answers from a bank at random in just 2 hours using UE4's Blueprints and UMG on a mobile device. I've played quiz games at museums before: they're usually about as fun as a job training quiz. However, the information can be very interesting, and there is potential there with a more dramatic approach. Mote knew full well how dramatic I was: I worked with my friend Meng Wang and his roommate Pat Vadanathorn on 2115, a short "video of a game" to teach others about Ocean Acidification. That video was inspired by Metroid Prime: set 100 years in the future in an underwater seascape, you are armed with a scanner to find information about objects in the world. Not a very traditional concept, but one that definitely appealed to judges and Mote staff to make a reality!
So, I didn't want this quiz to be traditional, either! I was a huge fan of Who Wants To Be A Millionaire growing up. I found a lot of strength in the classic answer questions, few lifelines, hard loss, no-gimmicks, no-nonsense kind of gameplay. But that still wasn't enough. A couple weeks of sitting on a project that seemed to be going nowhere, it happened in an instant, and something clicked: why not follow a manatee on a migration? I didn't quite have a word for the concept until I explained it to Evan and he immediately responded "Oh, it's like a journey!" The best publishers are the ones who know exactly what you want even before you do and are totally OK with it!
With the journey in place, the game had real motivation. You want to see what lies ahead and get to the end. 10 questions, 4 sections of difficulty. To attract guests initially the game runs a track mode cinematic showing off the different environments (a deep shelf, a jellyfish basin, a canal, and a mangrove forest). I made sure to show angles not usually seen in the game for the cinematic track mode and marketing materials to keep the gameplay experience unique.
Once we latched onto the journey concept I knew the sound I wanted to go for: a prominent piano with staccato strings and some orchestral backing (composed in MuLab 7 with Miroslav Philharmonik). I listened to Finding Nemo's sweeping orchestra (mainly Nemo Egg) and Who Wants to be a Millionaire's themes and progression for inspiration, but I didn't want the orchestra to sit in the background. The Game Over theme was the first one created: an upbeat melody played with somber notes and tempo.
Eventually I was able to figure out the Easy Questions theme, a much more active arpeggio, which became the base for all the other question themes. The violins play slowest on the easy questions and speed up as the quiz progresses, naturally making players feel more excited as the quiz progresses. The harp actually slows down by 1/3 for the middle questions, meeting the strings in the middle, the peak of "anxiety" in the quiz where losing becomes a definite possibility. It was easy to simply speed up the arpeggio for the hard questions. The electronic sweep was inspired by the sound of a car passing over a metal-grated drawbridge. The final question just needed to be the polar opposite: a very calm loop. The songs were made so they can interweave together during the quiz. Funnily enough, this may not have been necessary as the scene changes with the music.
Gameplay-wise I wanted this game to be challenging and rewarding. Playtesting proved quite successful as everyone from gamedev groups to co-workers enjoyed guessing the questions and moving forward in the quiz. The graphics were universally praised. The one negative I sometimes hear was with regards to the harshness of the Game Over and starting back at the beginning. Well, even before I had all the mechanics figured out, I knew I wanted that Game Over! There should be a sense of satisfaction and accomplishment for completing a difficult task, but that sense diminishes when the task is too easy. Of course this quiz can be beaten by anyone, given enough time to memorize the questions and get a more favorable question selection. I stick to my guns: there must be a lose state. I know not everyone will be able to make it to the end, and not everyone will love the Game Over, but I do believe they will love the game. That matters to me so much more than just the moment.
The most important feature was the wrong guess forgiveness: you can guess a 4-answer question incorrectly, find out it was wrong, and guess again. This game is filled with hope spots: spots with risky high stakes where you know you can fail, but you still have the confidence to progress and succeed. A straight game places a lot of pressure on just playing the game, which removes confidence, fun, and ultimately progress. The ability to fail and continue progressing keeps the game moving and adds a second tier between the fail and success state: a "caution" state, where the old "progress" state used to lie. This tiered dynamic means the player plays through the game alternating not between success and failure, but between success and caution, adding a layer of depth that was previously absent.
Visually I like my worlds to be vibrant, lively, and fun in the mundane. But it's never a lie. When I was a child, I went fishing on a boat in the inter-coastal waterways of Venice, FL and swimming/walking along beaches all over Sarasota County. A real underwater scene doesn't look so grandiose, but this is how I imagine Florida's gulf and inter-coastal waters would like to present themselves if they had a choice. I once had an experience at the beach where I was caught off-guard by a barely-visible oddly-shaped plastic bag in the water. It took me a second to realize I wasn't face-to-face with a sandwich bag!
Graphically, I wanted to shoot for the moon: 4K, 60f, HDR. The best possible specs available right now. And it wasn't easy: even with a GTX 1070, I had to optimize every shader I worked on to get the best results without killing the pixel budget. This can be seen nowhere better than the sand: since UE4's landscape is already tessellated, I bumped it up closer to the camera using custom code and quickly fade it out in the distance. That's real geometry for every bump of sand. In 4K, the rendering cost of pixel shading is 4x that of 1080p, so if an effect could be achieved through geometry, that would always be significantly preferred over pixel shaders... so I blew the vertex budget to get that sand and was gimped the rest of the project! That turned out to be well worth the hassle as I didn't need so many polygons in the distance anyways.
The grass movement used a new method I came up with called "Temporal Circular Displacement." Using sine/cosine waves to generate a circular motion and textures to offset that displacement, I can create the illusion of water currents ruffling sea grass around. Similar effects in math alone would've required significantly more vertex instructions. There is parallax occlusion on the concrete pillars, subsurface scattering on the sponges, and even a normal-based semi-metallic iridescent effect on all the fish, which wave their tails through standard parallax mapping. There is volumetric scattering in the water, actual refraction and translucency for the surface, and the whole world is lit by caustics. Not only are the caustics displaced by a panning normal map, but they also blur and darken towards the bottom of the ocean floor, simulating light scattering. While I made many optimizations in all these techniques (single texture displaced by normal map, blurring achieved via mip-biasing), this light is still far more complex than a solid directional term, bumping up the complexity of every single thing in the world. Between the volumetric fog, tesselation, refractive translucent surface, quality shaders, and caustics lighting, it is a miracle this project runs so effortlessly.
Optimizations were strong, but not fierce. Lower scalability settings were used for effects and shadows (High, not Epic). Instead of TXAA, I used FXAA and motion blur. Surprisingly, this did not look bad at all. Everything past the fog quickly drops LOD and culls out, especially the landscape. I had to decimate the heck out of my Mangrove roots: originally modeled with 60 thousand polygons, they never appear more than 8k in the game, and quickly get reduced to 2k. The water surface above is a grid mesh that calculates fog per-vertex. The concrete pillars have only 32 faces on a straight cylinder. There are no SSRs, just a few reflection probes, and culling is very aggressive. The fish are GPU sprites (and very convincing!). In probably one of my less inspired moves, I even took out cascaded shadow maps for the grass and relied entirely on screen-space shadowing. Just in an effort to maintain a solid 4K at 60f in HDR... with V-sync on. When your project is taking up a giant 65" TV in 4K, small issues with frame tearing or blurry resolutions become magnified many times over. This game might not be the most artistically complex masterpiece ever (most objects are very simple), but boy does it look good!
Overall, this Manatee Quiz game was the best professional project I've ever worked on. I had complete control: nothing appears in the game that I didn't make myself, and the final product ended up better than I have imagined it. It didn't start off as such a grand idea, but grew its ambition over time. It's a lot easier to dress up a simple idea than to decimate a complex one (KISS!). Nothing but good choices were made throughout, and I learned so much working on it. The game can be played today at Mote Marine Laboratories and Aquarium in the manatee exhibit!
PS - If you're up for it, there is a secret play mode hidden in this game. At any time during the quiz, hit the lifeline buttons left - center - right - center - left - right - left. Good Luck!...
No comments:
Post a Comment